Following up on “Feral”

Well, that one certainly got some attention. I’d posted it around 2 p.m. and then went out for a long walk to wind down, and came back about an hour later and saw there had been only a couple dozen views. Okay, so no one is interested…at least I wrote it. An hour after that, 700 views, then 1000, then 2000 [1]…finally leveled out around 7,500.[10][14]

Lots of nice notes from people—thanks—though it is interesting how different folks picked up on different bits. I don’t follow Twitter or PSJR [2] though I did follow back a few of the links from other blogs and thus have a couple—not seven—clarifications to make.

Yes, Virginia, I still believe in quantitative political analysis

There is a [hopeful] subset of commenters out there who have interpreted this as one of those end-of-career (or, more commonly, mid-career) denunciations of my prior quantitative work. Per, for example, Hayward Alker most famously, Richard Ashley or Ido Oren.

Wishful thinking, folks. For starters, yes, I am denouncing garbage can models—as did Achen a decade ago, and criticisms of this mis-use of regression modeling goes back decades before Achen—but I’ve never run garbage can models! I’m still doing quantitative work, and having left the constraints of academia, I’ll be doing even more of it than before. As I made very clear in the conclusion of 7DS—and in decided contrast to the position of Mearsheimer and Walt who feel that the quantitative curriculum should be dumbed down—I just think contemporary quantitative practice should be cleaned up.

What needs to be done is fairly clear and finite, and again, has been obvious for at least a decade:

  • Evaluate models according to out-of-sample accuracy—probably on multiple metrics—not significance levels;
  • Except in those circumstances—which are very rare in international and comparative research—where there is a good theoretical reason to expect a coefficient to be zero, don’t pay any attention to significance tests based on that null hypothesis;
  • If you want to assess the impact of a variable, do so using Bayesian methods and interpretations;
  • Use case-based control methods—we have a lot of these now—rather than garbage can “controls.”

Not really very complicated, right? In addition, my recommendation to someone just getting going would be:

  • Expand your skill set to include computational methods outside of classical statistics, and I’d now also suggest geospatial and advanced visualization methods;
  • Use the opportunities provided by the many new data sets and, more generally, the opportunities provided by the web for the production of near-real-time data

I’m walking the walk on these, not just telling others to do it. Not exactly a denunciation of quantitative analytical methods.

For those without an aptitude for statistics…

So not everyone wants to do statistics, or has an aptitude for it. I’m fine with that, and keep in mind I’ve done fieldwork and lived in five countries. [3] But if you aren’t going to learn statistics, put in a comparable amount of time doing one or more of the following

  • learn one or more languages other than your native language, preferably to a fairly high level of fluency rather than merely being able to do the “restaurant script.” Preferably one non-Indo-European language other than Finnish. Then spend at least six months living outside the country where you grew up. [4]
  • do extensive archival research.
  • spend significant time either in a policy position or working with people who are.

With one or more of those qualitative skills and experiences, you will probably have something interesting to say about politics. If you aren’t willing do any of these—and in particular, if you think all that is involved in studying politics is spending endless hours in windowless seminar rooms discussing Derrida and concepts beginning with “neo-“, I’m not interested in you, nor—let us be realistic—will you be of any  interest to most search committees. In contrast, if you invested the same sort of effort in serious qualitative work that we expect a methodologist to invest in quantitative training—and a surprising number of people now combine intensive qualitative field work with reasonable quantitative training—then I’m interested in what you have to say, and I’m guessing you’ll do fine in your career.

With a minor additional caveat in the event you are planning to write on “grand strategy”:

  • Don’t write anything until you are at least forty;
  • Don’t expect me to pay any attention to it until you’ve been dead at least fifty years.

It also helps if you’ve been a complete professional failure [5], or gotten yourself killed for your trouble. [6]


I gather there have been some interesting exchanges in the Twittersphere on this, and I may have tossed out a dangerously attractive concept that it will be even more attractive in six weeks when you are grading bluebooks. Two very important caveats

  • I’ve already been doing this for about twenty-five years and have both networks, and at the moment four independent income streams [7]
  • I have demonstrated technical skills

The first in particular takes quite a while—and no small amount of luck—to develop, and simply going out on your own early in a career—and I’m not really on my own, again, I’ve got long-established networks—is probably not advisable. So maybe I shouldn’t have made that quip about the attraction of being a [figurative] pirate.

That said, watching the recent experience of a couple young folks who have decided not to go the academic route, there is a real market for people with good quantitative skills in political analysis. In both instances, these folks found jobs fairly quickly, though both did, in fact, have to endure “How many golf balls fit in a 747?” hazing exercises in the interview process. [8] In one case, alas, the first job didn’t really work out, and the individual was forced to find alternative employment in one of the most attractive cities in the world at merely twice the salary of the first job.[11] So yeah, there are alternatives but—key caveat—both individuals had extensive multi-disciplinary technical skills, far beyond those of a basic political science methods sequence.

My guess, if you want to be a pirate, find a pirate ship, preferably with one or more senior folks on board, rather going out completely on your own. [9] In terms of U.S. government consulting, we’re obviously in a period of major downsizing—though I’ve also heard suggestions that this has eliminated a lot of marginal players, a good thing—but in terms of opportunities to make new use of data—“big” and otherwise—and computational power, we’re in yet another golden age.

Just a final thought, particularly as I move into a position where I may well be looking to hire people at some point: surely there is a more efficient way to produce people with these skills, right? That is, get Ph.D.-level technical training—which necessarily involves active involvement in research, more or less an apprenticeship—but without all of the other b.s. surrounding Ph.D. programs? I’m guessing there’s a solution out there—and in fact, it might look more like some combination of apprenticeship and MOOCs—but I’m not seeing it yet.

<13.08.08: Interesting take on this from a discussion this morning with someone also doing technical political analysis: “I need the sort of person who I can give a task to, and she comes back and says ‘Here’s the analysis you wanted, and here’s the analysis of the additional six variables you should have asked me about but didn’t.'” The question: where do you find such people?

Not, in my experience, the typical individual trained in engineering or computer science. Even if they’ve got an intuitive feel for politics—and most don’t [13]—they usually don’t have any sense of the existing state of the art or past research. So again, what we need are people who have enough qualitative knowledge of politics to build and assess plausible models, a basic knowledge of how quantitative political analysis is done and how to check the existing literature (e.g. from the conference paper archives that are not locked behind paywalls), and then a range of state of the art technical skills. That seems to me to involve about two years of graduate-level coursework, so something that looks more like an M.A. than a Ph.D. But I don’t really see that sort of program anywhere.<>

Cool places

Recommendations of cool places to consider living that are close to DC but outside the traffic

  • Winchester, VA
  • Haymarket, VA
  • Easton, MD
  • Poolesville, MD
  • Frederick, MD [15]
  • Shepherdstown, WV
  • Triangle area, NC

Thanks, and I’m still looking for more ideas.


[1] After three hours, the first rock through my living room window from an irate Penn State supporter…just kidding…

[2] Any more, though for a time several years ago…now it can be told…I was one of the dreaded anonymous moderators on PSJR. But after a while you get tired spending all of that time taking out the trash.

[3] My definition of “living in a country” is that you have to have rented an apartment and paid a local utility bill…you learn a lot about a country when you figure out how to pay a utility bill.

[4] And if that country was in North America or Europe, spend at least some of that time in a less-developed country where you will get really, really sick and thus will have incredibly gross and only slightly exaggerated stories to convey when you get drunk at the new student recruitment weekend. [12]

[5] Confucius, Thucydides, Machiavelli, Clausewitz, Marx

[6] Socrates, Hypatia, Jesus, Saul of Tarsus, Cicero, Seneca, Trotsky, Rosa Luxemburg, M.L. King

[7] And ample retirement savings, in part by virtue of not having children—Millennials, that’s you—who required extortionately expensive education at elite universities.

[8] Those sorts of interview questions, people are gradually realizing, are completely useless in predicting whether someone can do the job, but they make the interviewer feel smug and superior. Not, I’d suggest, dramatically different from the typical academic interview process.

[9] Sorry, at present I’m not hiring. Though I’ve been asked more than once.

[10] Almost the same response as I get to an APSR article. Joke. WordPress provides hourly stats and the curves on the number of views are both remarkably smooth—large samples—and quite consistent with a mixture of diffusion models initialized at multiple times. Probably a nice little article in there…

[11] The other was already living in one of the most attractive cities in the world, at least for those under 35.

[12] Hey, what about Australia? Dudes, every Australian I’ve known already has visited such places, acquired the experiences and—having downed a couple more beers than the rest of us—is likely to initiate the telling of the stories.

[13] Well, here. In the Middle East, a different story: at least where I’ve taught, the radical Islamic groups seem to disproportionately attract engineers. The liberal artsy types who we in the West tend to associate with radical movements, in contrast, tend to be secular social democrats.

[14] <13.08.09> Well no—still counting. Eight days past the original post, and we’re well beyond 18,000 views. I’d like to hope that with the impending start of the new academic year, at least a few of those got someone thinking that perhaps it is time to move along. Though what I’d really like to know is the number of time this has invoked, “Honey, I think you should read this…”

[15] Some differences of opinion here re: traffic.

This entry was posted in Higher Education, Methodology, The Blog. Bookmark the permalink.

2 Responses to Following up on “Feral”

  1. Pingback: Going Feral! Or “So long, and thanks for all the fish…” | asecondmouse

  2. Pingback: Feral Plus Three | asecondmouse

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s